From Alpha to Execution: Trading Horizons, Regimes, and Schedules under Concave Impact and Hurst Scaling
Abstract
Classical execution models assume linear impact, diffusive prices, and an exogenous horizon. We study execution under concave power-law impact and Hurst-scaled price fluctuations, and make two contributions. First, we choose the execution horizon by maximizing the Sharpe ratio of the trade's realized P&L. Price moves during the execution window are part of that P&L, so executing more slowly reduces its variance, and because impact is concave it can also lower cost per share. Urgency comes only from alpha realized while the order is being worked. The problem is governed by the signs of two quantities: (i) the duration elasticity of impact a, which under a structural prior equals the impact-persistence gap $β- H$ and can take either sign; and (ii) the net urgency ν, the rate of alpha realization net of exposure relief. When ν keeps its sign over the window, an interior optimum exists, to leading order in T/tf, only when both are positive; otherwise the order should be worked at the fastest or slowest admissible pace. In the interior regime the first-order condition involves no impact coefficient, order size or volume, and the objective is flat: for a large-cap example, the window within 1% of maximal Sharpe spans more than an order of magnitude. Second, given the window, we derive the schedule shape by convex programming under an envelope cost. For any increasing per-share cost of delay w(t), the KKT solution is ϕ∝ (λ− w(t))+1/β. For a Hurst-scaled envelope w ∝ tH, the normalized schedule depends only on $(β, H)$, has a closed form via the incomplete beta function, is front-loaded, and tapers smoothly to zero, saving 10-17% of envelope cost relative to a constant-rate (TWAP) schedule of equal duration. Whether it uses the whole window depends on a single dimensionless robustness level, for which we give a calibration.
Related
- The data geometry of masking diffusion: Certified-optimal schedules via unmasking growth complexity
- Accelerated Concatenation of Stepsize Schedules for Smooth Strongly Convex Gradient Descent
- Asymptotic Risk of Learning-Rate Schedules for Least-Squares Regression with Discretely Distributed Features
- Multiplicative vs. Additive State Coupling in Learned Optimizer Schedules
- Heavy-Ball Method under Randomized Schedules
BibTeX
@misc{kulkarni2026from,
title = {{From Alpha to Execution: Trading Horizons, Regimes, and Schedules under Concave Impact and Hurst Scaling}},
author = {Dnyanesh Kulkarni and Revant Nayar},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.23749},
eprint = {2610.23749},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.23749},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - From Alpha to Execution: Trading Horizons, Regimes, and Schedules under Concave Impact and Hurst Scaling
AU - Dnyanesh Kulkarni
AU - Revant Nayar
PY - 2026
DA - 2026/10/16
PB - arXiv
DO - 10.48550/arxiv.2610.23749
UR - https://arxiv.org/abs/2610.23749
AB - Classical execution models assume linear impact, diffusive prices, and an exogenous horizon. We study execution under concave power-law impact and Hurst-scaled price fluctuations, and make two contributions. First, we choose the execution horizon by maximizing the Sharpe ratio of the trade's realized P&L. Price moves during the execution window are part of that P&L, so executing more slowly reduces its variance, and because impact is concave it can also lower cost per share. Urgency comes only from alpha realized while the order is being worked. The problem is governed by the signs of two quantities: (i) the duration elasticity of impact a, which under a structural prior equals the impact-persistence gap $β- H$ and can take either sign; and (ii) the net urgency ν, the rate of alpha realization net of exposure relief. When ν keeps its sign over the window, an interior optimum exists, to leading order in T/t_f, only when both are positive; otherwise the order should be worked at the fastest or slowest admissible pace. In the interior regime the first-order condition involves no impact coefficient, order size or volume, and the objective is flat: for a large-cap example, the window within 1% of maximal Sharpe spans more than an order of magnitude. Second, given the window, we derive the schedule shape by convex programming under an envelope cost. For any increasing per-share cost of delay w(t), the KKT solution is ϕ∝ (λ− w(t))_+^1/β. For a Hurst-scaled envelope w ∝ t^H, the normalized schedule depends only on $(β, H)$, has a closed form via the incomplete beta function, is front-loaded, and tapers smoothly to zero, saving 10-17% of envelope cost relative to a constant-rate (TWAP) schedule of equal duration. Whether it uses the whole window depends on a single dimensionless robustness level, for which we give a calibration.
ER -